PCBRouteAgent: multimodal LLM agent routes 84/86 boards vs baseline 10/86
Yue Zhang · hf · 2026-11-10
PCBRouteAgent automates PCB routing with a multimodal LLM agent mirroring human engineers' global-strategy-plus-local-execution workflow:
- A Global Routing Planner uses board images plus structured electrical/geometric data to specify topology, routing order, corridors and layer transitions
- A Detailed Routing Executor realizes guides via KiCad's native routing engine with bounded local recovery
- Unresolved failures return as structured feedback enabling targeted guide revisions while preserving accepted routes
Evaluated on 247 synthetic and real boards across four PCBWorld-Bench splits, it hits clean-pass rates of 100%, 94.9%, 100% and 40% on D2, D3-A, D3-B and D3-C, beating the best baseline by 13.9, 36 and 10 points on the three real-world subsets. The figure shows 84/86 boards routed vs 10/86 for the baseline on one board.
More from coding & agent
- PSAISuite: A PowerShell Toolkit for Calling Multiple AI Models via One Interface — dfinke · 2026-11-10
- Eric S. Raymond shares his AI coding workflow: Threadripper rig, Codex, no autonomous commits — esrtweet · 2026-11-10
- Claude Code settings files expose thousands of secrets on GitHub — Ars Technica AI · 2026-11-10
- Chollet backs 'AI CEO' Steve that runs acquired SaaS firms with P&L as objective function — fchollet · 2026-11-10
- The most important dev skill in 2026: telling your agent "no, not like this" — dbreunig · 2026-11-10
- UC Berkeley open-sources FreeToken, an edge-native engine running frontier MoE models on consumer hardware — petrusenko_max · 2026-11-10